Nvidia to support gigawatt-scale computing infrastructure for SK, Naver

Nvidia partnered with Korean chipmaker SK hynix to develop next-generation memory chips that will power Nvidia's AI infrastructure. SK's telecommunication arm, SK Telecom, will begin operating its first AI factory in 2027 on Nvidia's platform.

The two major deals were announced Monday following a morning meeting between Nvidia CEO Jensen Huang and SK Group Chairman Chey Tae-won.

"SK is our largest memory partner," Huang told reporters after the meeting at SK's Seorin building in Jongno district, central Seoul. "We are expanding our partnership to include many new markets. [...] We announced a redesign, a reinvention of the world's personal computers we call RTX Spark, a partnership between us and Microsoft to reinvent the personal computer for the first time in 40 years, and that will have SK hynix inside.

"The next wave called physical AI and robotics — we built a processor called Jetson Thor, and that will have SK hynix inside."

The long-term partnership was struck to ensure a stable supply of advanced memory chips, which take years to design and bring to market, the companies said.

The deal is part of Nvidia's broader strategy to lock in formal partnerships across its entire semiconductor supply chain — the foundries that manufacture its chips, the memory suppliers, and the software companies that help design them.

Under the agreement, SK hynix and Nvidia will jointly develop memory for a wide range of Nvidia AI products, from data center supercomputers to consumer PCs and robotics systems — specifically the Vera Rubin AI supercomputer, Vera CPU, RTX Spark PC and Jetson Thor robotic computing platform.

SK hynix will also use Nvidia's software tools to speed up its internal chip development processes — work that traditionally requires enormous computing resources and time. This includes using Nvidia's CUDA-X software toolkit, which lets applications run much faster on Nvidia's chips, and PhysicsNeMo, an Nvidia framework that uses AI to simulate physical processes in chip manufacturing, such as how light interacts with a chip's surface during production.

The partnership will extend to electronic design automation — the software used to design chips — to form a three-way collaboration between chipmakers, Nvidia and the companies that make chip design software.

The two companies are also working together on digital twin technology to enable real-time AI monitoring and smarter decision-making on the production floor.

For SK Telecom, the focus is on building out AI factories: purpose-built facilities optimized for AI workloads, going beyond what conventional data centers can do. The first is expected to go online in Korea next year, built on Nvidia's DSX platform — a blueprint that covers everything from the chips inside to the software, power infrastructure, and operational systems.

"AI factories are essential for Korea's universities, scientific labs, startups and industries," Huang said. "Just like electricity, water and the internet, Korea will be powered by AI in the future. It will be used in every country, every company and every industry — including, of course, the manufacturing of chips and telecommunications."

SK Telecom will join Nvidia's Cloud Partner program, an ecosystem of companies that deliver AI computing services using Nvidia's infrastructure, with the goal of offering competitive pricing and energy efficiency.

SK Group and Nvidia will also launch a joint research & development (R&D) effort on AI factory architecture, establishing a working body focused on improving how GPUs and memory chips are designed to work together from the ground up.

"Through our close partnership with Nvidia, we have built full-stack AI infrastructure competitiveness spanning chips through data center operations," Chey said. "Beyond service delivery, we will jointly tackle challenges around GPUs, memory and energy — and emerge as a leading AI cloud provider driving AI ecosystem development across Asia."

The U.S. chip giant also unveiled a cooperation with Naver, the country’s top internet firm cited by Huang as a key partner during Computex 2026, to build a gigawatt-level computing infrastructure dubbed AI Factory.

In the beginning phase, the Korean firm plans to operate a 55-megawatt facility in the first half of 2027, and then expand the infrastructure capacity to 100 megawatts within the same year and 200 megawatts by 2028 before reaching the ultimate gigawatt level.

Naver Chairman Lee Hae-jin said the alliance was "highly encouraging" because it would allow the company to present "a concrete alternative" for countries and regions around the world seeking to establish their own sovereign AI capabilities.

"The partnership is particularly meaningful as it creates an opportunity for Naver's technological and infrastructure strengths to expand to the global stage," he said.

With machinery to construction conglomerate Doosan, Nvidia will extend the area of cooperation across physical AI, robotics and AI infrastructure with an aim of integrating Nvidia’s advanced AI platforms into systems of Doosan affiliates.

For instance, Doosan Robotics is adopting robotics frameworks called Isaac Sim and Isaac Lab and robot-focused platforms such as Jetson Thor and Agentic Robot OS, while Doosan Enerbility is exploring the use of DSX, an architectural blueprint for building large-scale data centers. 

BY LEE JAE-LIM, PARK EUN-JEE [[email protected]]